Lucio Dona' wrote:
When the boards are powered up without Ethernet cable plugged-in, and
the Ethernet is initialized, it seems that the system gets a lot of
'rx interrupts' and stays almost all the time inside the Ethernet rx
interrupt service routine.
This horribly slows down everything and causes the watchdog to come up
resetting the board.
Tried to make some debugging, but don't understand why there are all
If the system is powered up with the Ethernet cable connected to a
simple hub (hub only, no other computers connected to the net) the
problem does not occur.
I have this problem on about 10% boards.
The design is based on AMD Syrah reference schematics, the kernel is
2.4.21-pre4. Searching with oscilloscope there is no apparent
'hardware' difference between a 'bad' board and a good one.
If this is on 10% of the boards I would say it is a production problem.
What PHY do you use and is it properly reseted?
Is there any soldering problem with the PHY. Does it have a ground pad
under the case and is this pad well soldered?